How to understand causation?

  • Find A is causal, rather than random, consequence.
  • How do we identify the causes of A?
  • How do we justify B is a cause of A?

Pluralist view?

  • \(\neq\) multicausality

Pluralist view: formal causes, material causes, efficient causes, deterministic causes, probabilistic causes, correlational causation, causal mechanisms…

  • Different explanations for the same causation
  • Different perspective of the same causal explanation.

Why not pluralist view?

  • Over-stating the different-ness
  • Not benefiting the research

A universal view

  • A minimal definition
  • 16 criteria of formal properties of causal arguments
  • 8 criteria for research design

Defining causation

Cause: Events or conditions that raise the prior probability of some outcome occurring, under ceteris paribus conditions (Gerring 2005, 169).

  • \(P(Y|X) > P(Y|-X).\)
  • Why a minimal causation?
    • Hint: Sartori's ladder
  • Bayesian framework?
    • \(Y(A|B) = \frac{Y(B|A)Y(A)}{Y(B)}\)
    • \(Posterior = \frac{Likelihood\times Prior}{Evidence}\)

Causal Proposition

  • Specification
  • Precision
  • Breadth
  • Boundedness
  • Completeness
  • Parsimony
  • Differentiation (exogeneity)
  • Priority
  • Independence
  • Contingency
  • Mechanism
  • Analytic utility
  • Intelligibility
  • Relevance
  • Innovation
  • Comparison

Criteria of Demonstration

  1. Plenitude
  2. Comparability
  3. Independence
  4. Representativeness
  5. Variation
  6. Transparency
  7. Replicability

Plenitude

Conducting an empirical based study.

Nope

Yes

Comparability

  • Descriptive comparability: 'X' and 'Y' mean roughly the same thing across cases.
  • Causal comparability: X and Y do not interact in idiosyncratic ways in different cases.
  • Control: the extent to which remaining dissimilarities (of both sorts) may be taken into account.

Independence and Representativeness

Variation

Transparency

Replicability

Two strategies to test theory

  • Actual case strategy (save for later)
  • Counterfactual strategy

What's counterfactual?

  • Claims about events that did not actually occur.

Relation with hypothesis test?

  • Following the experimental logic

  • Compromising with the reality

Differences from the hypothesis test

Hypothesis test

  • Rely on "ceteris paribus"
  • Some probability assumptions
  • Can assess the frequencies and magnitudes of the causality
  • Uncertainty can be reduced by more cases

Counterfactual

  • Rely on general principles, laws, or regularities
  • Knowledge of historical facts
  • Assess effects based on proliferation
  • No formal criterion of uncertainty

Why not actual cases?

  • Comparability
  • Degree of freedom

When to use?

  • Qualitative, mostly
  • # of variables > # of observations

Risk

How can we know what would have happened?